Gibberellin regulates Arabidopsis floral development via suppression of DELLA protein function

Transcription

1 Research article 1055 Gibberellin regulates Arabidopsis floral development via suppression of DELLA protein function Hui Cheng 1,2, *, Lianju Qin 3, *, Sorcheng Lee 1, *, Xiangdong Fu 4, Donald E. Richards 4, Dongni Cao 1, Da Luo 3,, Nicholas P. Harberd 4, and Jinrong Peng 1,2, 1 Institute of Molecular & Cell Biology, 30 Medical Drive, Singapore Department of Biological Sciences, National University of Singapore, Singapore Institute of Plant Physiology & Ecology, 300 Fenglin Road, Shanghai , China 4 John Innes Centre, Colney Lane, Norwich NR4 7UH, UK *These authors contributed equally to this work Authors for correspondence ( dluo@sibs.ac.cn, harberd@bbsrc.ac.uk and pengjr@imcb.a-star.edu.sg) Accepted 18 November 2003 Development 131, Published by The Company of Biologists 2004 doi: /dev Summary The phytohormone gibberellin (GA) regulates the development and fertility of Arabidopsis flowers. The mature flowers of GA-deficient mutant plants typically exhibit reduced elongation growth of petals and stamens. In addition, GA-deficiency blocks anther development, resulting in male sterility. Previous analyses have shown that GA promotes the elongation of plant organs by opposing the function of the DELLA proteins, a family of nuclear growth repressors. However, it was not clear that the DELLA proteins are involved in the GA-regulation of stamen and anther development. We show that GA regulates cell elongation rather than cell division during Arabidopsis stamen filament elongation. In addition, GA regulates the cellular developmental pathway of anthers leading from microspore to mature pollen grain. Genetic analysis shows that the Arabidopsis DELLA proteins RGA and RGL2 jointly repress petal, stamen and anther development in GA-deficient plants, and that this function is enhanced by RGL1 activity. GA thus promotes Arabidopsis petal, stamen and anther development by opposing the function of the DELLA proteins RGA, RGL1 and RGL2. Key words: Gibberellin, DELLA proteins, Stamen development, Floral development Introduction Gibberellin (GA) is an endogenous regulator of plant growth (Richards et al., 2001). Severely GA-deficient Arabidopsis mutants, such as ga1-3, exhibit retarded vegetative growth of shoots (Koornneef and van der Veen, 1980; Peng and Harberd, 1997; King et al., 2001) and roots (Fu and Harberd, 2003). In addition, the development of floral organs, especially petals and stamens, is impaired in GA-deficient mutants, while retarded anther development results in male-sterility owing to a lack of mature pollen (Wilson et al., 1992; Goto and Pharis, 1999). All of the GA-deficient mutant floral phenotypes of ga1-3 can be restored to normal by application of exogenous GA (Koornneef and van der Veen, 1980), demonstrating that GA regulates floral development (Pharis and King, 1985). Anther development is GA dependent in a range of plant species additional to Arabidopsis (Nester and Zeevaart, 1988; Jacobsen and Olszewski, 1991), suggesting that GA is a general regulator of floral development. The histology of anther development (microsporogenesis) is well documented (Goldberg et al., 1993; McCormick, 1993; Sanders et al., 1999). In addition, the study of mutants exhibiting stage-specific defects in microsporogenesis and pollen release has further advanced our understanding of the process (Preuss et al., 1993; Sanders et al., 1999; Park and Twell, 2001; Azumi et al., 2002; Steiner-Lange et al., 2003; Unte et al., 2003). Although little is known about how GA controls stamen filament elongation, anther development or microsporogenesis, there have been previous suggestions that GA-signalling components may modulate these processes. For example, overexpression of SPY (an Arabidopsis GAsignalling component) (Jacobsen et al., 1996) inhibits postmeiotic anther development in petunia (Izhaki et al., 2002). Furthermore, transgenic expression of wild-type or mutant forms of GAI (another Arabidopsis GA-signalling component, see below) can retard stamen elongation and induce malesterility in tobacco and Arabidopsis (Huang et al., 2003; Hynes et al., 2003). By contrast, infertility caused by impaired floral development is also a characteristic of mutants lacking the rice or barley DELLA proteins SLR1 or SLN1 (Ikeda et al., 2001; Chandler et al., 2002). Despite these various reports, the mechanism via which GA regulates petal, stamen and anther development remained unclear. Recent advances have enabled the identification of a family of proteins homologous to Arabidopsis GAI and RGA (Peng et al., 1997; Silverstone et al., 1998) that are crucial for the regulation of plant stem elongation growth in response to GA (Peng et al., 1999; Ikeda et al., 2001; Boss and Thomas, 2002; Chandler et al., 2002). These proteins belong to the DELLA family (Fleck et al., 2002), a subfamily of the GRAS family of putative transcriptional regulators (Pysh et al., 1999; Richards

2 1056 Development 131 (5) Research article et al., 2000). The Arabidopsis genome encodes five distinct DELLA proteins (Lee et al., 2002). Genetic suppression studies have shown that GAI and RGA functions overlap in the repression of plant stem growth (Dill and Sun, 2001; King et al., 2001). Further studies showed that while RGL2 controls seed germination (Lee et al., 2002), RGL1 may control stem elongation as well as seed germination (Wen and Chang, 2002). Although GAI, RGA, RGL1 and RGL2 are all expressed in developing inflorescences (Lee et al., 2002), no obvious suppression of ga1-3 floral phenotype was observed in ga1-3 mutants lacking GAI, RGA, GAI and RGA, or RGL2 (Dill and Sun, 2001; King et al., 2001; Lee et al., 2002). However, a transgenic RGL1 loss-of-function line was resistant to the arrest of floral organ development induced by paclobutrazol (PAC, an inhibitor of GA biosynthesis) (Wen and Chang, 2002), suggesting that RGL1 might play a role in regulating floral development. These observations underscore the importance of determining systematically the respective roles of the various DELLA proteins in GA-mediated regulation of Arabidopsis petal and stamen development. In this report, we describe experiments addressing two questions. First, we characterized ga1-3 floral development to determine at which stage of stamen/anther development GAdeficiency causes developmental arrest. Second, we used novel combinations of loss-of-function mutations to determine if DELLA proteins are repressors of stamen filament elongation and microsporogenesis. Our results show that GA is crucial both for cell elongation during stamen elongation and for the developmental progression from microspore to mature pollen grain during pollen development. We also show that the DELLA proteins RGA, RGL1 and RGL2 work together to repress stamen and anther development in GA-deficient plants. Materials and methods Plant growth conditions, genetic nomenclature, plant materials Plants were grown as described previously (Lee at al., 2002), or in controlled long (LD)- and short (SD)-day growth conditions (Peng and Harberd, 1997). Genotypes are written in capital italics (e.g GAI), mutant genotypes in lowercase (e.g. gai-t6). Polypeptide gene products are written in nonitalic capitals (e.g. GAI). Mutant lines (Landsberg erecta background) rgl1-1, rgl2-1, gai-t6, rga-t2 and ga1-3 were as described (Peng et al., 2002; Lee et al., 2002). The ga1-3 lines lacking various combinations of RGL1, RGL2, GAI or RGA were as described previously (King et al., 2001; Lee et al., 2002), or made via cross-pollination. Primer pairs used for genotype verification were as described previously (Lee et al., 2002). For anatomical analysis of ga1-3, seeds were chilled for 7 days, following which the seed coat was manually removed and the seed sown on soil. Histology and in situ hybridization For scanning electron microscopy, individual flower buds from fresh wild-type or mutant inflorescences were dissected; outer organs were removed using stainless steel needles. Buds were attached to a mounting plate, plunged into liquid nitrogen, quickly transferred to a specimen chamber and scanned at 10 KV (JSM-5310LV, JEOL, Japan). Pollen grains were mounted on scanning electron microscopy stubs and coated with gold using previously described techniques (Bozzola and Russell, 1999). For anther sectioning, fresh inflorescences were fixed in formalin-acetic acid-alcohol (FAA) fixative buffer at 4 C overnight followed by dehydration steps and subsequent embedding in Jung Historesin (Leica). Sections (2.5 µm) were made using a Leica RM 2055 microtome and stained with 0.25% Toluidine Blue O (Sigma). DAPI staining of pollen grain nuclei was performed as described (Chen and McCormick, 1996) and pollen numbers were counted under a microscope (Leica DM RXA2) with 40 or 20 objectives. Color photos were taken using a Spot Insight QE digital camera (Diagnostic Instruments). The ATA7 and SDS antisense and sense probes were synthesized from the pmc1577 and pmc2317 plasmids, respectively (Zhao et al., 2002) and in situ hybridisation was performed as described previously (Luo et al., 1996). Callose staining and chromosome spread analysis of meiotic stages were as described (Regan and Moffatt, 1990; Ross et al., 1996). Results GA regulates epidermal cell elongation during stamen elongation To determine the developmental stage at which ga1-3 flower buds become arrested, the relative growth of stamens versus gynoecium in inflorescence flower buds [starting with the outmost (oldest) and ending with the innermost dissectable bud] was compared in ga1-3 and wild type via scanning electron microscopy (SEM). All floral organs were properly initiated in ga1-3, and appeared to develop normally, though only to around floral stage 10 (Fig. 1A) [floral stages as defined elsewhere (Smyth et al., 1990; Bowman, 1994)]. Subsequent petal and stamen development beyond stage 10 was arrested in ga1-3 (compare wild type at floral stage 13 with ga1-3; Fig. 1A). The ultimate stamens and pistils of ga1-3 flower buds were much shorter than the mature stamens and pistils of the wild-type control (Fig. 1B). In addition, ga1-3 had a stronger effect on stamen filament length than on pistil length: the arrest of stamen development resulted in significantly shorter stamens versus pistils in ga1-3 (Fig. 1B). SEM of stamen filament epidermal cells indicated that the arrest of stamen filament growth in ga1-3 was due to reduced cell length (Fig. 1C), rather than to a reduction in cell number (Fig. 1D). ga1-3 plants fail to produce tricellular pollen grains In Arabidopsis, the anther consists of four lobes, each of identical cell-type architecture, each derived from archesporial cells. Microsporogenesis initiates within the reproductive locule of each lobe. The sporogenous cells divide to generate microspore mother cells (MMC). Subsequently, the MMC undergo meiosis to generate tetrads of haploid microspores (MSp). The MSp are released from the tetrads and undergo two rounds of cell division to form mature tricellular pollen grains (Sanders et al., 1999). The surface structure of the mature pollen grains released by wild-type plants was compared with that of immature pollen grains manually dissected from ga1-3 anther locules. All wild-type pollen grains were oval shaped with long indented lines. Very few oval shaped pollen grains were observed in ga1-3. Instead, in most cases, the immature pollen grains from ga1-3 plants were spherical in shape (Fig. 2A). Wild-type and ga1-3 anthers were dissected and stained with DAPI. As expected, the mature pollen grains from wild-type plants were tricellular, and contained three nuclei (Fig. 2B). However, in most cases (see Discussion), fewer than 10% of the developing grains examined in ga1-3 pollen sacs were found to be bicellular/tricellular (Fig. 2B,C). In fact, about 48% of ga1-3 pollen grains contained only a single nucleus

3 DELLA proteins regulate floral development 1057 Fig. 1. GA regulates stamen filament length via control of cell elongation. (A) SEMs of wild-type flowers at floral stages 10 and 13 are shown. Petals and stamens from the most advanced flower (middle bottom) in ga1-3 resembled the wild-type flower at stage 10 (top left). Stamen filament elongation is dramatically increased in ga1-3 plants lacking RGL1, RGL2 and RGA (Q2) and in ga1-3 plants lacking RGL1, RGL2, GAI and RGA (penta) compared with the wild type. Both Q2 and penta lines produced visible pollen grains. (B) Comparison of stamen and pistil lengths among different genotypes. Filament and pistil lengths were measured from SEM pictures (n=20). (C) SEM of stamen filament epidermal cells. ga1-3 stamen filaments have relatively short epidermal cells compared with those of wild type; cell length was restored in stamen filaments of ga1-3 plants lacking RGL1, RGL2, GAI and RGA (penta). Stamen filament segments shown were all from the middle part of the filament. (D) Average number of epidermal cells per stamen filament in wild type, ga1-3 and ga1-3 lacking RGL1, RGL2, GAI and RGA (penta). and 46% had no nucleus. Clearly, ga1-3 fails to produce mature pollen, and this probably results from an arrest or impairment in pollen development prior to or during pollen mitosis in ga1-3 (Fig. 2C). Microsporogenesis is arrested before pollen mitosis in ga1-3 As in other plant species, microsporogenesis in Arabidopsis is a highly programmed process, and can be divided into 14 stages at the cellular level (Sanders et al., 1999). In wild type, floral stage 10 corresponds to stage 7-8 of microsporogenesis (pre-mitosis) (Bowman, 1994). As shown above, petal and stamen growth arrests at around floral stage 10 in ga1-3, suggesting that microsporogenesis may arrest prior to pollen mitosis in this mutant. To define the specific stage of microsporogenesis arrested in the ga1-3 mutant, we compared transverse sections of anthers from ga1-3 and wild type. No obvious differences were observed between ga1-3 and wild type up to microsporogenesis stage 7, when tetrads are formed (Fig. 3). However, after stage 7, it appeared that the microspores in ga1-3 were not properly released and that the ga1-3 pollen sacs failed to expand. In addition, the ga1-3 tapetum remained at the vacuole stage and then subsequently degenerated together with the microspores and pollen sacs (Fig. 3). As a result, the later stages of microsporogenesis (stages 9-12) could not be convincingly distinguished in ga1-3, suggesting that microsporogenesis is arrested at stages 7-8 (prior to pollen mitosis), thus preventing the formation of mature pollen. To confirm that microsporogenesis was arrested prior to pollen mitosis in ga1-3, we used the gene expression markers SDS and ATA7. SDS is a marker gene that is specifically expressed in microsporocytes (Azumi et al., 2002). In situ hybridization analysis revealed near-identical expression patterns for SDS in the anthers of ga1-3 mutants and wild type, suggesting that the early stages of microsporogenesis are not affected in ga1-3 (Fig. 4A). In fact, chromosome spread experiments confirmed that meiosis in ga1-3 was successfully accomplished, resulting in the formation of tetrads (Fig. 4B). However, Aniline Blue staining showed that, although the callose wall was formed successfully, the tetrads in ga1-3 anthers were not properly separated from one another, as is seen in wild type (Fig. 4C). ATA7 is an early tapetum-specific molecular marker (Rubinelli, 1998). As expected, ATA7 was highly expressed in wild type between stages 7 and 9 (Fig. 4A). The ATA7 signal then gradually disappeared during the later stages of microsporogenesis, although pollen mitosis was occurring (data not shown). ATA7 was also strongly expressed in ga1-3 anthers (Fig. 4A), implying that tapetum initiation is not detectably affected by GA deficiency. However, in most cases, the ga1-3 pollen sacs marked by ATA7 were irregular in shape and contained microspores that had become deformed and severely degenerate at a developmental stage prior to the disappearance of the ATA7 signal (Fig. 4A).

4 1058 Development 131 (5) Research article Fig. 2. ga1-3 plants fail to produce tricellular pollen grains. (A) Pollen grains from various genotypes examined by SEM. Pollen grains from ga1-3 plants lacking RGL1, RGL2 and RGA (Q2), or RGL2 and RGA (not shown) were almost identical to wild-type pollen grains, being oval shaped with long indented lines on the surface. In ga1-3 plants, the cells contained in a locule appeared to remain as microspores, being round and, in some cases, with short indented lines on the wall surface. Pollen grains from ga1-3 plants lacking RGL1, RGL2, GAI and RGA (penta) were similar to wild-type pollen grains but were slightly more wrinkled in appearance. (B) DAPI staining showed that pollen grains from ga1-3 plants lacking RGL1, RGL2 and RGA (Q2), or RGL1, RGL2, GAI and RGA (penta) are tricellular (as in the wild-type control). By contrast, the majority of the cells in the anther locule of ga1-3 had either no detectable nucleus or only a single condensed nucleus whereas only ~6% had two or more than two nuclei. (C) Frequencies of tricellular pollen grains in anther locules of various genotypes as revealed by DAPI staining as shown in B. Absence of specific DELLA combinations suppresses ga1-3 floral phenotype To investigate if DELLA proteins are repressors of floral development, we studied floral development in ga1-3 plants lacking various combinations of GAI, RGA, RGL1 and RGL2. First, we studied ga1-3 plants lacking single DELLA proteins. Absence of RGL1, RGL2 or GAI alone had little obvious effect on the phenotype of ga1-3, while absence of RGA alone partially suppressed the stem elongation phenotype of ga1-3 (Silverstone et al., 1998) (Fig. 5A). However, absence of RGA alone did not restore normal floral development or fertility to ga1-3 (Fig. 5C). Second, we studied ga1-3 plants lacking all possible pairwise combinations of RGL1, RGL2, GAI or RGA. Lack of GAI and RGA completely suppressed the dwarf phenotype conferred by ga1-3 (Dill and Sun, 2001; King et al., 2001) (Fig. 5B). All other combinations caused a phenotype that was indistinguishable from that of ga1-3 (plants lacking RGL1 and RGL2, or lacking RGL1 and GAI, or lacking RGL2 and GAI) or from ga1-3 plants lacking RGA (plants lacking RGL1 and RGA, or lacking RGL2 and RGA). Thus, GAI and RGA in combination play the predominant role in controlling stem growth. Absence of RGL1 and RGL2 did not obviously enhance the suppression of ga1-3 phenotype because of absence of RGA, suggesting that RGL1 and RGL2 have relatively minor roles in the regulation of stem elongation. For the most part, the pair-wise DELLA absence combinations failed to confer normal flower development on ga1-3 (Dill and Sun, 2001; King et al., 2001). All of these lines produced flower buds, but the buds failed to open and exhibited the arrested petal and stamen growth characteristic of ga1-3 (Fig. 7A). However, some flower opening was observed in the late maturity of two of the pair-wise DELLA absence combination lines. Although flowers of 40-day old ga1-3 plants lacking RGL2 and RGA were sterile, at ~50 days and older these plants produced flowers that opened and were able to set seed (Fig. 5C). In addition, the flowers of late maturity ga1-3 plants lacking GAI and RGA (~55 days old; data not shown) sometimes opened, but these flowers were almost always sterile. RGL1, RGL2 and RGA act synergistically in the repression of Arabidopsis stamen and petal development We next studied ga1-3 plants lacking all possible three-way combinations of RGL1, RGL2, GAI and RGA. Absence of RGL1, RGL2 and GAI failed to suppress any detectable aspect of ga1-3 phenotype (Fig. 6A,B and Fig. 7A). Absence of

5 DELLA proteins regulate floral development 1059 Fig. 3. Histological analysis of microsporogenesis. Transverse sections of anthers are displayed in developmental sequence, showing the progress in microsporogenesis in various genotypes. ga1-3 anthers developed normally up to the tetrad formation stage (stage 7) but after this, they diverted from the normal (compared with stages 9-12 in wild type; ga1-3 stages highlighted with question marks). Eventually, all ga1-3 pollen sacs aborted (last two photos). ga1-3 plants lacking RGL1, RGL2 and RGA (Q2), or RGL1, RGL2, GAI and RGA (penta) successfully completed microsporogenesis and released mature viable pollen grains. Scale bar: 50 µm. Microsporogenesis stages are indicated in bottom right-hand corner of the pictures. RGL1, RGL2 and RGA completely restored petal and stamen development to ga1-3 (Fig. 1A, Fig. 7A), and permitted normal seed set (Fig. 6B, Fig. 7B), despite the fact that this line was semi-dwarf and exhibited a stem elongation phenotype only slightly taller than that of ga1-3 plants lacking RGA alone (compare Fig. 5A with Fig. 6B). Although ga1-3 plants lacking RGL2 and RGA produced fertile flowers only in late maturity, ga1-3 plants lacking RGL1, RGL2 and RGA produced fertile flowers at the onset of flowering. Thus RGL1, RGL2 and RGA act in combination to control petal and stamen development in response to GA. ga1-3 plants lacking RGL1, GAI and RGA were taller than ga1-3 plants lacking GAI and RGA alone (Fig. 6A,B), suggesting that RGL1 has a significant role in the control of stem elongation when RGA and GAI are absent. However, lack of RGL1, GAI and RGA did not restore normal stamen and petal development to ga1-3 (Fig. 1B, Fig. 7A), and this line was therefore sterile (Fig. 7B). In fact, the young flower Fig. 4. Pollen development is arrested in ga1-3. (A) The antherspecific markers SDS and ATA7 were used in in situ hybridization analysis to compare microsporogenesis in ga1-3 and wild type. SDS patterns appeared the same in ga1-3 and wild type whereas the ATA7 signal pattern in ga1-3 locules was significantly different to that of the wild-type control. (B) Chromosome spread experiments confirmed that pollen meiosis is successfully completed in ga1-3, resulting in tetrad formation. (C) Aniline Blue staining showed that the tetrads in ga1-3 tend to be clustered and are not found in the form of free microspores as is seen in the wild type. Scale bars: 1 µm in B; 50 µm in C. buds of ga1-3 plants lacking RGL1, GAI and RGA were indistinguishable from those of ga1-3 plants lacking GAI and RGA only (Fig. 7A). ga1-3 plants lacking RGL2, GAI and RGA were also taller at maturity than control lines lacking GAI and RGA alone (Fig. 6A,B,D). In contrast to what was seen with lack of RGL1, lack of RGL2 (in ga1-3 plants lacking RGL2, GAI and RGA) partially restored petal and stamen development to ga1-3 plants lacking GAI and RGA, making this line partially fertile (Fig. 1B and Fig. 7A,B). In summary, the results described in this section indicate that GA-regulation of Arabidopsis petal and stamen elongation is mediated via RGL1, RGL2 and RGA, with RGL2 and RGA playing the predominant roles. Absence of RGA, RGL2, RGL1 and GAI leads to GAindependent plant growth Finally, we analysed ga1-3 plants lacking RGL1, RGL2, GAI and RGA. We found that this mutant line bolted and flowered

6 1060 Development 131 (5) Research article Fig. 5. RGA and RGL2 are key GA-response regulators of floral development. (A,B) Comparison of growth of ga1-3 plants lacking single (A) or pairs of (B) DELLA proteins at 48 days. Presence of wild-type gene (and functional protein) is indicated by +, presence of loss-of-function mutation (and lack of functional protein) is indicated by. (C) ga1-3 plants lacking RGA and RGL2 initially produced sterile non-opening flowers (bottom right; SEM far right), then began to produce fertile open flowers (top right; SEM far right) when they were ~50 days old. By contrast, ga1-3 plants lacking RGA alone (middle) never produced any fertile flowers. earlier than wild type both in long (LD) (Fig. 6A,C) and short (SD) days (Fig. 6C). Furthermore, ga1-3 plants lacking RGL1, RGL2, GAI and RGA were taller than the wild-type control (Fig. 6B,D). In addition, combined absence of RGL1, RGL2, GAI and RGA suppressed the effects of ga1-3 on petal and stamen development. The flowers of ga1-3 plants lacking RGL1, RGL2, GAI and RGA exhibited fully extended stamens and petals (Fig. 7A). Anther development proceeded to completion, resulting in flowers that were fertile and set seeds in both LD (Fig. 7B) and SD (data not shown). We next analyzed stamen filament growth in ga1-3 plants lacking RGL1, RGL2, GAI and RGA and found that these filaments were slightly longer than those of the wild-type control (Fig. 1A,B). In fact, removing only RGL1, RGL2 and RGA from ga1-3 also resulted in filament lengths longer than that of wild type (Fig. 1A,B). SEM of stamen filament epidermal cells indicated that restoration of stamen filament length in ga1-3 mutant plants lacking RGL1, RGL2, GAI and RGA was due to an increase in cell length (Fig. 1C) as opposed to an increase in cell number (Fig. 1D), a difference similar to what was previously observed between wild-type and ga1-3 stamen filaments. Thus, the elongation of stamen filaments becomes GA independent when all four DELLA proteins are removed. SEM analysis of pollen grains from ga1-3 plants lacking RGA, RGL2 and RGL1 showed that their surface structure was indistinguishable from those of wild type (Fig. 2A), being oval-shaped with long indented lines. However, pollen grains from ga1-3 plants lacking RGL1, RGL2, GAI and RGA were substantially different from those of wild type, mostly having a wrinkled appearance, and sometimes being markedly deformed (Fig. 2A). Although ga1-3 plants lacking RGL1, RGL2 and GAI had hugely different stem elongation phenotypes to ga1-3 plants lacking RGL1, GAI and RGA (Fig. 6A,B), the floral phenotypes of these two lines are very similar and both lines had ~30% tricellular pollen (Fig. 2C). This suggests that microsporogenesis is partially restored in these two lines. Lack of both RGL2 and RGA had a greater effect on microsporogenesis, as ~60% or ~80% of pollen grains were found to be tricellular in ga1-3 plants lacking RGL2, GAI and RGA or RGL1, RGL2 and RGA respectively (Fig. 2B,C). These results indicate that RGA and RGL2 play important roles in the repression of microsporogenesis in Arabidopsis, and that GA regulates microsporogenesis by overcoming the repressing effects of RGA and RGL2. Transverse sectioning showed that ga1-3 plants lacking RGL1, RGL2 and RGA or ga1-3 plants lacking RGL1, RGL2, GAI and RGA both achieved complete microsporogenesis (Fig. 3). However, although no obvious differences were observed between wild-type and ga1-3 plants lacking RGL1, RGL2 and RGA, we often observed that one or two of the four locules of the anthers of ga1-3 plants lacking RGL1, RGL2, GAI and RGA were aborted (data not shown). Discussion DELLA proteins act as repressors of growth whose function is opposed by GA (Richards et al., 2001). In several cases,

7 DELLA proteins regulate floral development 1061 Fig. 6. Absence of RGL1, RGL2, GAI and RGA leads to GA-independent plant growth. (A,B) Wild-type plants were compared with ga1-3 plants containing loss-of-function mutations causing lack of various combinations of RGL1, RGL2, GAI or RGA at 22 (A) and 48 (B) days. Note that ga1-3 plants lacking RGL1, GAI and RGA or RGL2, GAI and RGA or all four DELLA proteins bolted earlier than the wild-type control at 22 days. Genotypes are as indicated. (C) Flowering time of various genotypes, measured as number of leaves at time of flowering. In this case, ga1-3 plants under SD did not produce any visible flower buds even though they had more than 50 leaves (marked with *). Under LD condition, ga1-3 plants lacking RGL1, GAI and RGA or RGL2, GAI and RGA or all four DELLA proteins all flowered with one or two leaves less than did the wild-type control, although this small magnitude of difference is hard to discern from the histogram shown here. (D) Plant height of various genotypes at 28 days old under LD growth condition. Results are presented as mean±standard error (n=30). GA opposes DELLA function by promoting DELLA disappearance, most likely via ubiquitin-ligase dependent targetting to the proteasome and subsequent protein degradation (Silverstone et al., 2001; Ito et al., 2002; Gubler et al., 2002; Fu et al., 2002). Recent studies have identified candidate F-box components of a SCF E3 ubiquitin ligase apparently responsible for targetting DELLA proteins to the proteasome (McGinnis et al., 2003; Sasaki et al., 2003). Thus, one emerging picture is that GA stimulates GA-responses by targetting DELLA protein growth repressors for destruction in the proteasome (Harberd, 2003). In another case, that of RGL2, GA apparently opposes RGL2 function by causing downregulation of RGL2 transcript levels during seed germination (Lee et al., 2002). Previous studies of Arabidopsis DELLA function have involved phenotypic comparisons of GA-deficient (ga1-3) plants with GA-deficient plants lacking GAI, RGA, RGL1, or RGL2 or a limited range of combinations of these factors (Dill and Sun, 2001; King et al., 2001; Lee et al., 2002). In this paper, we described the effects of a more comprehensive set of DELLA-lack combinations, focussing especially on floral development. Flowering consists of three distinct phases: floral initiation (in which the vegetative meristem is transformed into an inflorescence meristem), floral organ initiation and floral organ growth. As shown previously, lack of GAI and RGA substantially suppresses the effect of the ga1-3 mutation on flowering time (a measure of time of floral initiation) in SD (Dill and Sun, 2001). We have shown that an additional lack of RGL2 or of both RGL1 and RGL2 further advances the flowering time (in both LD and SD) of ga1-3 plants lacking GAI and RGA. However, the magnitude of this further advance is relatively small compared with that initially caused by lack of both GAI and RGA. Thus, GAI and RGA play the predominant role in regulating flowering time in the GAsignalling floral promotive pathway (Simpson and Dean, 2002), with only small contributions from RGL1 and RGL2. By contrast, RGL1, RGL2 and RGA play key roles in floral organ development. The temporal coordination of the development of individual floral organs is essential for floral function. For example, at around the time that the pollen matures and is dehisced from the anther, the stamen filaments of flowers of self-fertilizing species such as Arabidopsis elongate and bring the pollen into contact with the stigmatic papillae (Smyth et al., 1990; Bowman, 1994). We showed that the relatively short stamen filaments of ga1-3 flowers result from an arrest of cell elongation rather division and that combined lack of RGL1, RGL2, GAI and RGA restored stamen filament cell elongation in ga1-3 plants. We also showed that, in general, microspores do not proceed to the formation of mature pollen in ga1-3 anthers, that microspore development is possibly arrested prior to pollen mitosis in ga1-3, and that tapetal development is perturbed in ga1-3. Whether the effect of ga1-3 on pollen mitosis is a secondary effect of arrested tapetal development, or is independent of the effect on tapetal development is at present unclear. In addition, we occasionally observed ga1-3 flower buds containing a significant number of tricellular pollen grains. Further investigation is needed to find out if this is a true reflection of the ga1-3 developmental process or is caused by other unknown environmental cues. Lack of RGL1, RGL2, GAI and RGA proteins restored microsporogenesis in ga1-3 plants. Further genetic analysis enabled us to identify RGL2, RGA and RGL1 as the key GA-response regulators controlling stamen filament length and microsporogenesis. Interestingly, pollen grains from ga1-3 plants lacking GAI, RGL1, RGL2 and RGA, although tricellular and viable, are deformed when compared with the wild-type-appearing pollen grains from

8 1062 Development 131 (5) Research article Fig. 7. RGL1, RGL2 and RGA repress flower opening, petal and stamen development in ga1-3 plants. (A) Comparison of the flowers of 30-dayold plants; genotypes as indicated. (B) Comparison of the seed set of various genotypes. (Left panel) Segment of main shoots bearing siliques. (Right panel) Seed production in a typical silique from genotypes as shown. Although ga1-3 plants lacking RGL1, GAI and RGA, or RGL2, GAI and RGA both have short siliques, the former bears no seeds whereas the latter is partially fertile. Siliques of ga1-3 plants lacking RGL1, RGL2 and RGA or lacking all four DELLA proteins are similar not only in length but seed production as well as in wild type. Scale bar: 1.0 mm. ga1-3 plants lacking RGL1, RGL2 and RGA. Perhaps absence of all four DELLA proteins activates the GA pathway to such high levels that pollen wall materials are overproduced, resulting in abnormal pollen morphology. Previous developmental genetic analyses showed that the Arabidopsis DELLA proteins GAI and RGA act as repressors of stem elongation and that GA exerts its promotive effects on stem growth by overcoming the effects of GAI and RGA (Dill and Sun, 2001; King et al., 2001). These observations, and additional observations on the behaviour of DELLA proteins in other species, have been incorporated into a general release of restraint model, which envisages DELLA proteins as general agents of restraint of plant organ growth, and GA as a means of overcoming that restraint (Peng et al., 1997; King et al., 2001; Richards et al., 2001; Harberd, 2003). However, the initial experiments (which examined the effect of lack of Arabidopsis GAI and RGA) showed that although stem elongation could be explained in terms of the release of restraint model, other aspects of growth and development which were known to be GA regulated (in particular seed germination and floral organ growth) could not (Dill and Sun, 2001; King et al., 2001). It therefore remained possible that some other, entirely different, mechanism was responsible for the GA-mediated regulation of seed germination and floral organ growth. Recently, it has been reported that the GA-promotion of Arabidopsis seed germination can be explained in terms of a GA-mediated release of the restraint upon germination imposed by RGL2 (Lee et al., 2002) or RGL1 (Wen and Chang, 2002). The results in this present paper show for the first time that the GA-regulation of floral organ development is also DELLAmediated. However, different combinations of DELLA proteins are key to floral organ development (RGA, RGL1, RGL2), seed germination (RGL2 and RGL1) and stem elongation (RGA, GAI). The three key aspects of the ga1-3 mutant phenotype (dwarfism, inhibition of seed germination, retarded floral organ development) can now be explained: the lack of GA in this mutant causes a failure to overcome the repressive effects of the DELLA protein combinations that are specific to each particular phenotypic aspect. As a consequence, the release of DELLA restraint hypothesis can now be considered to be a viable model with which to understand GA responses in general. One possible explanation for how different DELLA combinations control different developmental processes (e.g. seed germination versus stem elongation versus stamen development) is that individual DELLA proteins have different temporal and spatial expression patterns. For example, GAI and RGA are ubiquitously expressed in all plant tissues, whereas RGL1 and RGL2 transcripts are relatively enriched in the inflorescence (Silverstone et al., 1998; Lee et al., 2002; Wen and Chang, 2002). In situ hybridisation showed that RGL1 is highly expressed in the stamen primordium (Wen and Chang, 2002); however examination of an RGL2 promoter-gus fusion line showed that RGL2 transcripts are also enriched in the stamen (Lee et al., 2002). The expression patterns of RGL1 and RGL2 are therefore consistent with our current observation that RGL1 and RGL2 are both important for stamen development. It is possible that RGL3 also plays an important role in various aspects of GA-mediated developmental regulation. Determination of the relative role of RGL3 awaits

9 DELLA proteins regulate floral development 1063 characterization of Arabidopsis mutants lacking the functional RGL3 allele. The nature of the arrest in flower development conferred by ga1-3 (and restored by lack of RGL1, RGL2 and RGA) is particularly interesting. Our results identify a relatively distinct developmental stage at which arrest occurs. Before that stage, ga1-3 stamen and anther development proceeds in a way that is indistinguishable from that of wild type. After that stage, wild-type development continues, while ga1-3 development is blocked. It will be interesting to determine if other GAdeficiency phenotypes (e.g. the particular shape of leaves of ga1-3 mutant plants) are also due to premature arrest of an identifiable developmental sequence. Plasmids pmc1577 (ATA7) and pmc2317 (SDS) were kindly provided by Dr Hong Ma. We are grateful for Yang Sun Chan and Qing Wen Lin for their help with SEM; and De Ye, Weicai Yang, Lin Guo and Chen Jun for advice on techniques and useful discussions. We thank Zoe Wilson for comments on the manuscript. This work is financially supported by the Agency for Science, Technology and Research in Singapore. The work of X.F., D.R. and N.P.H. was supported by the Core Strategic Grant from the Biotechnology and Biological Sciences Research Council to the John Innes Centre and by Biotechnology and Biological Sciences Research Council Grant 208/P The work of L.Q. and D.L. was supported by grants from the National Natural Science Foundation of China ( ) and the Chinese Academy of Sciences (KSCX2-SW-309). References Azumi, Y., Liu, D., Zhao, D. Z., Li, W., Wang, G. F., Hu, Y. and Ma, H. (2002). Homolog interaction during meiotic prophase I in Arabidopsis requires the SOLO DANCERS gene encoding a novel cyclin-like protein. EMBO J. 21, Boss, P. K. and Thomas, M. R. (2002). Association of dwarfism and floral induction with a grape green revolution mutation. Nature 416, Bowman, J. (1994). Arabidopsis: An Atlas of Morphology and Development. New York: Springer-Verlag. Bozzola, J. J. and Russell, L. D. (1999). In Electron Microscopy: Principles and Techniques for Biologists, pp Sudbury, MA: John and Bartlett. Chandler, P. M., Marion-Poll, A., Ellis, M. and Gubler, F. (2002). Mutants at the Slender1 locus of Himalaya barley: molecular and physiological characterization. Plant Physiol. 129, Chen, Y. S. and McCormick, S. (1996). Sidecar pollen, an Arabidopsis thaliana male gametophytic mutant with aberrant cell divisions during pollen development. Development 122, Dill, A. and Sun, T.-p. (2001). Synergistic derepression of gibberellin signaling by removing RGA and GAI function in Arabidopsis thaliana. Genetics 159, Fleck, B. and Harberd, N. P. (2002). Evidence that the Arabidopsis nuclear gibberellin signaling protein GAI is not destabilized by gibberellin. Plant J. 32, Fu, X., Richards, D. E., Ait-Ali, T., Hynes, L. W., Ougham, H., Peng, J. R. and Harberd, N. P. (2002). Gibberellin-Mediated Proteasome-Dependent Degradation of the Barley DELLA Protein SLN1 Repressor. Plant Cell 14, Fu, X. and Harberd, N. P. (2003). Auxin promotes Arabidopsis root growth by modulating gibberellin response. Nature 421, Goldberg, R. B., Beals, T. P. and Sanders, P. M. (1993). Anther development: basic principle and practical applications. Plant Cell 5, Goto, N. and Pharis, R. P. (1999). Role of gibberellin in the development of floral organs of the gibberellin-deficient mutant, ga1-1, of Arabidopsis thaliana. Can J. Bot. 77, Gubler, F., Chandler, P., White, R., Llewellyn, D. and Jacobsen, J. (2002). GA signalling in barley aleurone cells: control of SLN1 and GAMYB expression. Plant Physiol. 129, Harberd, N. P. (2003). Relieving DELLA restraint. Science 299, Huang, S., Cerny, R. E., Qi, Y. L., Bhat, D., Aydt, C. M., Hanson, D. D., Malloy, K. P. and Ness, L. A. (2003). Transgenic studies on the involvement of cytokinin and gibberellin in male development. Plant Physiol. 131, Hynes, L. W., Peng, J., Richards, D. E. and Harberd, N. P. (2003). Transgenic expression of the Arabidopsis DELLA proteins GAI and gai confers altered gibberellin response in tobacco. Transgenic Res. 12, Ikeda, A., Ueguchi-Tanaka, M., Sonoda, Y., Kitano, H., Koshioka, M., Futsuhara, Y., Matsuoka, M. and Yamaguchi, J. (2001). Slender rice, a constitutive gibberellin response mutant, is caused by a null mutation of the SLR1 gene, an ortholog of the height-regulating gene GAI/RGA/RHT/D8. Plant Cell 13, Itoh, H., Ueguchi-Tanaka, M., Sato, Y., Ashikari, M. and Matsuoka, M. (2002). The gibberellin signaling pathway is regulated by the appearance and disappearance of SLENDER RICE1 in nuclei. Plant Cell 14, Izhaki, A., Borochov, A., Zamski, E. and Weiss, D. (2002). Gibberellin regulates post-microsporogenesis processes in petunia anthers. Physiol. Plant. 115, Jacobsen, S. E. and Olszewski, N. E. (1991). Characterization of the arrest in anther development associated with gibberellin deficiency of the gib-1 mutant of tomato. Plant Physiol. 97, Jacobsen, S. E., Binkowski, K. A. and Olszewski, N. E. (1996). SPINDLY, a tetratricopetide repeat protein involved in gibberellin signal transduction in Arabidopsis. Proc. Natl. Acad. Sci. USA 93, King, K. E., Moritz, T. and Harberd, N. P. (2001). Gibberellins are not required for stem growth in Arabidopsis thaliana in the absence of GAI and RGA. Genetics 159, Koornneef, M. and van der Veen, J. H. (1980). Induction and analysis of gibberellin sensitive mutants in Arabidopsis thaliana (L.) Heynh. Theor. Appl. Genet. 58, Lee, S. C., Cheng, H., King, K. E., Wang, W., He, Y., Hussain, A., Lo, J., Harberd, N. P. and Peng, J. R. (2002). Gibberellin regulates Arabidopsis seed germination via RGL2, a GAI/RGA-like gene whose expression is upregulated following imbibition. Genes Dev. 16, Luo, D., Carpenter, R., Vincent, C., Copsey, L. and Coen, E. (1996). Origin of floral asymmetry in Antirrhinum. Nature 383, McCormick, S. (1993). Male gametophyte development. Plant Cell 5, McGinnis, K. M., Thomas, S. G., Soule, J. D., Strader, L. C., Zale, J. M., Sun, T. P. and Steber, C. M. (2003). The Arabidopsis SLEEPY gene encodes a putative F-box subunit of an SCF E3 ubiquitin ligase. Plant Cell 15, Nester, J. E. and Zeevaart, J. A. D. (1988). Flower development in normal tomato and a gibberellin-deficient (ga-2) mutant. Am. J. Bot. 75, Park, S. K. and Twell, D. (2001). Novel patterns of ectopic cell plate growth and lipid body distribution in the Arabidopsis Gemini pollens mutant. Plant Physiol. 126, Peng, J. R., Carol, P., Richards, D. E., King, K. E., Cowling, R. J., Murphy, G. P. and Harberd, N. P. (1997). The Arabidopsis GAI gene defines a signaling pathway that negatively regulates gibberellin responses. Genes Dev. 11, Peng, J. R. and Harberd, N. P. (1997). Gibberellin deficiency and response mutations suppress the stem elongation phenotype of phytochrome-deficient mutants of Arabidopsis. Plant Physiol. 113, Peng, J. R., Richards, D. E., Hartley, N. M., Murphy, G. P., Devos, K. M., Flintham, J. E., Beales, J., Fish, L. J., Worland, A. J., Pelica, F. et al. (1999). Green Revolution genes encode mutant gibberellin response modulators. Nature 400, Peng, J. R., Richards, D. E., Moritz, T., Ezura, H., Carol, P. and Harberd, N. P. (2002). Molecular and physiological characterization of Arabidopsis GAI alleles obtained in targeted Ds-tagging experiments. Planta 214, Pharis, R. P. and King, R. W. (1985). Gibberellins and reproductive development in seed plants. Annu. Rev. Plant Physiol. 36, Preuss, D., Lemieux, B., Yen, G. and Davis, R. W. (1993). A conditional sterile mutation eliminates surface components from Arabidopsis pollen and disrupts cell signaling during fertilization. Genes Dev. 7, Pysh, L. D., Wysocka-Diller, J. W., Camilleri, C., Bouchez, D. and Benfey, P. N. (1999). The GRAS family in Arabidopsis: sequence characterization and basic expression analysis of the SCARECROW-LIKE genes. Plant J. 18, Regan, S. M. and Moffatt, B. A. (1990). Cytochemical analysis of pollen development in wild-type Aarabidopsis and a male-sterile mutant. Plant Cell 2,

10 1064 Development 131 (5) Research article Richards, D. E., Peng, J. and Harberd, N. P. (2000). Plant GRAS and metazoan STATs: one family? BioEssays 22, Richards, D. E., King, K. E., Ait-ali, T. and Harberd, N. P. (2001). How gibberellin regulates plant growth and development: a molecular genetic analysis of gibberellin signaling. Annu. Rev. Plant Physiol. Plant Mol. Biol. 52, Ross, K. J., Fransz, P. and Jones, G. H. (1996). A light microscopic atlas of meiosis in Arabidopsis thaliana. Chromosome Res. 4, Rubinelli, P., Hu, Y. and Ma, H. (1998). Identification, sequence analysis and expression studies of novel anther-specific genes of Arabidopsis thaliana. Plant Mol. Biol. 37, Sanders, P. M., Bui, A. Q., Weterings, K., McIntire, K. N., Hsu, Y. C., Lee, P. Y., Truong, M. T., Beals, T. P. and Goldberg, R. B. (1999). Anther developmental defects in Arabidopsis thaliana male-sterile mutants. Sex. Plant Reprod. 11, Sasaki, A., Itoh, H., Gomi, K., Ueguchi-Tanaka, M., Ishiyama, K., Kobayashi, M., Jeong, D. H., An, G., Kiitano, H., Ashikari, M. and Matsuoka, M. (2003). Accumulation of phosphorylated repressors for gibberellin signaling in an F-box mutant. Science 299, Silverstone, A. L., Ciampaglio, C. N. and Sun, T.-p. (1998). The Arabidopsis RGA gene encodes a transcriptional regulator repressing the gibberellin signal transduction pathway. Plant Cell 10, Silverstone, A. L., Jung, H.-S., Dill, A., Kawaide, H., Kamiya, Y. and Sun, T.-p. (2001). Repressing a repressor: gibberellin-induced rapid reduction of the RGA protein in Arabidopsis. Plant Cell 13, Simpson, G. G. and Dean, C. (2002). Arabidopsis, the Rosetta stone of flowering time? Science 296, Smyth, D. R., Bowman, J. L. and Meyerowitz, E. M. (1990). Early flower development in Arabidopsis. Plant Cell 2, Steiner-Lange, S., Unte, U. S., Eckstein, L., Yang, C., Wilson, Z. A., Schmelzer, E., Dekker, K. and Saedler, H. (2003). Disruption of Arabidopsis thaliana MYB26 results in male sterility due to non-dehiscent anthers. Plant J. 34, Unte, U. S., Sorensen, A. M., Pesaresi, P., Gandikota, M., Leister, D., Saedler, H. and Huijser, P. (2003). SPL8, an SBP-box gene that affects pollen sac development in Arabidopsis. Plant Cell 15, Wen, C. K. and Chang, C. (2002). Arabidopsis RGL1 encodes a negative regulator of gibberellin responses. Plant Cell 14, Wilson, R., Heckman, J. W. and Somerville, C. (1992). Gibberellin is required for flowering in Arabidopsis thaliana under short days. Plant Physiol. 100, Zhao, D. Z., Wang, G. F., Speal, B. and Ma, H. (2002). The EXCESS MICROSPOROCYTES1 gene encodes a putative leucine-rich repeat receptor protein kinase that controls somatic and reproductive cell fates in the Arabidopsis anther. Genes Dev. 16,